Investigation of the Fingernail Plate for Biometric Authentication using Deep Neural Networks


Biometric traits from the dorsal part of the hand have been subjected to limited investigation, as compared to traits from the palmer part. However, these limited studies have established the dorsal traits as biometrics of immense potential. Significant biometric traits from the dorsal part of the hand that have proved their worth in biometric authentication include geometry of the fingers [1], knuckles [2-5], and hand vein thermograms [6,7]. The fingernail plate, another biometric trait belonging to the dorsal part of the hand, has only been explored very recently for personal authentication [8]. The extremely minuscular research, that the fingernail plate has been subjected to, has provided reasonably appreciable results.

Motivation and Scope of Present Work

While the nail plate regenerates with the creation of new cells, the spacing amidst the grooves found in the nail bed (as shown in Figure 9.1) has been found to remain proportionally constant throughout the life of a person [9]. Also, nail-ridge patterns illustrate a high extent of individuality, even in the case of identical twins [10].

It is of importance to note that potential trace evidences found in crime investigation sites often include photographs (as shown in Figure 9.2.) and videos of the hand dorsum. These type of evidences which encompass one or more fingernail plate(s) can be suitably processed to match with samples acquired from the person suspected of being the criminal.

A major disadvantage of popular biometrics like fingerprints and palmprints is that people leave these marks on whatever they touch. This renders a biometric system involving one or both of these two traits very susceptible to impersonation. This is not the case with nail plates, and as such these cannot be impersonated very easily. As such, authors have chosen this trait and subjected it to exhaustive investigation in this work, so as to examine the adequacy of the trait in biometric authentication.

Owing to the better performance of multibiometric systems [11,12], and aiming towards enhanced performance, the current work has been largely dedicated to a multimodal design, where one fingernail plate has been subjected to fusion with one or more fingernail plates. Even in case of criminal investigations, trace evidences

Anatomy of Human Fingernail

FIGURE 9.1 Anatomy of Human Fingernail.

Samples of Trace Evidences encompassing the Fingernail Plate

FIGURE 9.2 Samples of Trace Evidences encompassing the Fingernail Plate.

are more likely to comprise more than one fingernail plate, as seen in Figure 9.2. Also, nail damage caused by infection [8,13] are likely to reduce the achievable accuracy. Multimodal systems shall largely make up for such adverse results. All possibilities have been considered, and rigorous experiments have been performed to address all such situations and to investigate the nail plate in a multimodal biometric setup. A framework of the overall processing scheme of the proposed work has been provided in Figure 9.3.


The preliminary and pioneering works carried out using the fingernail involved the usage of very heavy sensors [14,15]. A variety of equipment like transmitted light comparison microscope, cross polarising filters equipped for polarised light, and acrylic resin were required for the flawless acquisition of nail samples [14]. For sampling, the study carried out in Ref. [15] required a whole range of apparatus including an acousto-optic 2D beam deflector, thermoelectric coolers, master oscillator, a pair of highly monochromatic light sources, and photodiode array or charge-coupled device (CCD) sensor.

Framework of the Overall Processing Scheme of the Proposed System

FIGURE 9.3 Framework of the Overall Processing Scheme of the Proposed System.

As per the best of the authors’ knowledge, the very first work that threw light on the fingernail plate as a biometric trait [16] introduced the Region of Interest (ROI) extraction technique used in this work. However, the reported technique did not remove the grown nail region, a part that provides nothing but redundant information. The mentioned work used low resolution white light for the sample acquisition of fingernail images, Haar Wavelet as the feature extraction technique, and it investigated verification systems. The mentioned work did not explore identification systems. The next work in this domain [17] reported a method to remove the grown nail part. It used Haar Wavelet and Independent Component Analysis (ICA) as the feature extraction techniques. The said work mainly investigated the performance of the nail plate in verification systems; and only explored the same in identification systems at a very preliminary level. Also, the designing of the multibiometric systems adopted in this work was not robust enough. The subsequent work that checked the performance of nail plates in personal authentication of individuals [8] also used Haar Wavelet and ICA as the feature extraction algorithms. It reported designing a multibiometric setup as well, and exploited only a couple of preliminary fusion techniques. However, the limited experimentation gave appreciable results. The first work [5] that has carried out exhaustive experiments under a deep learning framework using a multibiometric fusion of the nail plate and the knuckle reported notable results. This study has used deep learning features exploiting the AlexNet model, and has established the nail plate as a potential biometric identifier in both verification and identification modes. The very promising results reported in the multimodal identification mode opened up avenues to further investigate the nail plate in the light of personal authentication.

The current study explores the fingernail plate further as a biometric trait in personal authentication, and also serves as a counter-narrative to the situation when the framework of Ref. [5] cannot be adopted, which may be caused due to the non-availability of usable and/or acceptable finger knuckle images. The primary objective of this work is to design efficient multimodal systems using the nail plate to check for better efficacy in personal authentication; and also to explore the nail plate under a broader spectrum of deep learning models. A major advantage that any system using the fingernail plate enjoys is its limited chance of impersonation, primarily because the traces of nail plate are not left by any person on anything that she/he touches during day-to-day activities.

< Prev   CONTENTS   Source   Next >